white paper for china unicom's edge-cloud platform architecture … · figure 2.1 edge...
TRANSCRIPT
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
I Copyright©China Unicom Network Technology Research Institute, 2018
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
Contents
1 OVERVIEW ................................................................................................................... 1
1.1 Vision of the White Paper ................................................................................................... 1
1.2 Version of the White Paper ................................................................................................. 2
2 SERVICE DEMAND AND NETWORK EVOLUTION TREND ................................ 3
2.1 Service Demand................................................................................................................... 3
2.2 Network Cloudification Evolution ...................................................................................... 5
3 EDGE-CLOUD PLATFORM ARCHITECTURE AND EVOLUTION ...................... 7
3.1 Overall Layout of Communication Cloud .......................................................................... 7
3.2 Edge-Cloud Platform Architecture ..................................................................................... 9
3.2.1 Infrastructure Architecture ....................................................................................... 9
3.2.2 Virtualized Layer Architecture ................................................................................ 13
3.2.3 Service Orchestration and Management Architecture ............................................ 16
3.2.4 Edge-Cloud Platform Capabilities ........................................................................... 18
3.3 Evolution Roadmap of the Edge-Cloud Platform............................................................. 26
4 EDGE-CLOUD STANDARDIZATION AND INDUSTRIAL CHAIN ...................... 29
4.1 Standardization Progress .................................................................................................. 29
4.2 Situation of the Industrial Chain ...................................................................................... 30
4.3 Orientation of China Unicom's Edge Cloud ..................................................................... 32
5 KEY SERVICE SCENARIOS AND COMMERIAL APPLICATIONS .................... 33
5.1 Scenario 1: Private Customization of Campus Area Network ......................................... 33
5.2 Scenario 2: Mobile vCDN ................................................................................................. 34
5.3 Scenario 3: Video Monitoring and AI Analysis ................................................................. 35
5.4 Scenario 4: Augmented Reality (AR) ................................................................................ 37
5.5 Scenario 5: Live VR Broadcast ......................................................................................... 38
5.6 Scenario 6: IoV C-V2X ..................................................................................................... 39
5.7 Scenario 7: Indoor Positioning .......................................................................................... 41
5.8 Scenario 8: Industrial IoT ................................................................................................. 43
6 SUMMARY AND FORECASTING ............................................................................ 45
CHINA UNICOM'S EDGE-CLOUD ECOLOGICAL PARTNERS ............................... 46
ABBREVIATIONS ........................................................................................................... 47
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
1 Copyright©China Unicom Network Technology Research Institute, 2018
White Paper for China Unicom's Edge-Cloud Platform
Architecture and Industrial Eco-System
1 Overview
1.1 Vision of the White Paper
As the information technology deeply changes our daily life, digital reshaping will
penetrate into all industries. The communication industry has undergone several leaps
and reforms during the past decades. From dial-up Internet access to FTTH and from
GSM/TD-SCDMA/Wimax to LTE, information pipes have undergone radical changes,
and the IT cloudification technology is reshaping CT and OT industries. The
architecture of both the wireless network and wired network has been evolving into
C/U separation and Fixed Mobile Convergence (FMC) to cope with all challenges
brought by industrial digital transformation and industrial structure upgrade. In the
predictable future, the integration at the infrastructure layer is imperative.
With the combination of increasingly mature SDN/NFV, big data, and artificial
intelligence, 5G networks will become key infrastructure in the digital transformation
of all industries. In 5G-oriented IoE scenarios, new types of services have the
characteristics of a lower latency, larger bandwidth, and more intelligence. The
traditional vertical type network architecture has many deficiencies in resource sharing,
agile innovation, flexible expansion, and simple O&M. To effectively meet the service
requrements of eMBB, mMTC, and uRLLC and enhance industrial competitiveness,
global telecom operators successfively begin network reconstruction and
transformation, aiming to construct a DC-centered full cloudification network.
The multi-access edge computing technology is the result of ICT integration, and
is also a key technology that allows telecom operators to pursue 5G-oriented network
transformation. This meets the development requirements of HD video, VR/AR,
industrial IoT, and IoV in the future. Operators also regard tens of thousands of edge
DCs as the best high-quality resources compared with OTT, gaining a wide application
space for edge computing. As the official communication service partner for the 2022
Winter Olympics, China Unicom has already established an Edge-Cloud strategic
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
2 Copyright©China Unicom Network Technology Research Institute, 2018
partnership with BAT, and has begun the planning and construction of thousands of
edge DCs. China Unicom is committed to building an open and open-source Edge-
Cloud Service PaaS Platform and providing rich platform service capabilities and
unified APIs for application developers, aiming to accelerate the incubation and
commercial use of innovative services related to edge applications.
As the pilot project that is constructed based on the edge ecology incubation and
edge service platform, China Unicom works with ZTE, Intel, and Tencent video to build
an Edge DC testbed in Jingjin University Town, Tianjin. The deployment tests of edge
vCDNs, edge transcoding, and edge intelligent analysis are already completed. For the
edge vCDN, the deployment of applications to the network edge not only reduces the
bandwidth pressure and latencies (compare to the traditional CDN mode, the average
RTT latency is reduced by 50 percent, and the HTTP download rate is increased by 43
percent) but also significantly enhances user experience and helps content providers
reduce costs. This pilot project blazes a trail in the edge computing collaboration
between China Unicom and OTT, and is of great guidance value for the reconstruction
of the edge DCs and the commercial incubation of 5G innovative services.
This white paper describes the architecture and evolution roadmaps of China
Unicom's Edge-Cloud platform in accordance with 5G FMC requirements and network
cloudification evolution. It also detailedly describes China Unicom's key service
scenarios and business cases that are based on the Edge-Cloud platform, with a
combination of the standardization progress of the edge computing technology and the
status of the industrial chain.
1.2 Version of the White Paper
In June 2017, China Unicom issued the White Paper for China Unicom’s Edge
Computing Technology at Shanghai MWC. This white paper further extends the Edge-
Cloud platform architecture, evolution planning, the advancement of commercial use,
and ecosystem construction, and hopes to be helpful for the development of the whole
industrial chain. With the pre-commercial deployment of 5G networks, freezing of
ETSI and 3GPP standards, more contents will be added to the future version, and any
comment or suggestion is welcomed.
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
3 Copyright©China Unicom Network Technology Research Institute, 2018
2 Service Demand and Network Evolution Trend
Throughout the history of telecommunications, every industry reform, from wired
to wireless, from voice to video, is driven by service demand. In addition, the reform in
an industry may be driven by the technological development of other industries. For
example, the photography industry began with film but now is reshaped by digital
images. The development of edge computing is driven by both of the following factors:
service demand and cloud-enabled evolution.
2.1 Service Demand
With the popularization of 4G and optical fiber communication networks, people
have become accustomed to the convenience brought by broadband services and
abundant applications, and desire more realistic and interactive experience. At the same
time, the rapid development of the Internet of Things (IoT) makes the originally isolated
equipment interconnected with each other and improves efficiency. Therefore, the
future information society will be bound to present diverse demand. Following the trend
of the time, the 3GPP defines three scenarios based on the research on 5G mobile
networks, namely Enhance Mobile Broadband (eMBB), Massive Machine Type of
Communication (mMTC), and Ultra Reliable Low Latency Communication (uRLLC).
• 5G eMBB provides users with ultra HD videos and immersive service
experience (such as VR and AR) through a transmission rate of 10 Gbps.
• The mMTC technology supports the access and interconnection of massive
devices typically in smart cities and smart buildings by connecting millions of devices
per square kilometer.
• By virtue of the ultra-low latency and ultra-high reliability, uRLLC greatly
improves the operation efficiency of the applications in vertical industries, such as the
Internet of Vehicles and industrial Internet.
Meanwhile, the development of fixed-network services also shows a similar 5G
trend. For example, the 8K video, 3D video, industrial control, and government and
enterprise private cloud services are evolving toward high bandwidth, low latency, and
massive connection. According to the development trends of wireless and fixed
networks, operators will face the following service demand challenges:
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
4 Copyright©China Unicom Network Technology Research Institute, 2018
• The eMBB services, such as 8K videos, 3D videos, and VR/AR, will create a
demand for ultra-high network bandwidth (hundreds of Gbps), thus causing great
transmission load on backhaul networks. The investment in capacity expansion of
convergence networks and Metropolitan Area Networks (MANs) can significantly
improve the transmission costs per media stream, but cannot generate investment
income.
• The URLLC services, typically V2X and industrial Internet, require an end-
to-end ultra-low latency of 1 ms. This cannot be fulfilled just through the technology
progress in the physical layer and transport layer of the wireless and fixed networks.
Therefore, it is necessary to introduce innovative deployment of network and industry
applications according to the characteristics of vertical industries.
• The mMTC services, such as smart city and smart building, will produce a
large amount of data, leading to huge challenges to operation management. Practices
show that only the centralized cloud cannot afford such a complex system. Instead,
localized intelligent control and management must be introduced.
Figure 2.1 Edge Computing Supports Fixed-Mobile Service Development
From the perspective of operators, it is a great challenge to fully support the service
scenarios above without decreasing investment income. As shown in Figure 2.1, we
believe that effectively use of storage, computing, network, and acceleration resources
provided by the edge service platform can satisfy the performance requirements for the
development of future 5G and fixed broadband services, which sinks the edge of some
key service applications to the edge of access networks. This can reduce thebandwidth
and latency loss caused by network transmission and multi-level forwarding, and
decrease the CAPEX and OPEX through the localized management and operation on
the unified edge computing platform.
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
5 Copyright©China Unicom Network Technology Research Institute, 2018
2.2 Network Cloudification Evolution
In order to satisfy the development requirement of future services in the high
bandwidth, low latency, and local management, the control plane is separated from the
forwarding plane in the 3GPP 4G CUPS and 5G New Core, thus implementing a flat
network architecture. The forwarding plane gateway is migrated to the wireless side,
distributed as required, and scheduled by the control plane in a centralized manner.
Combined with edge computing, gateway anchors can support end-to-end massive
services with low latency, high bandwidth, and balanced load, fundamentally resolving
the overload and latency bottlenecks in transmission and core networks caused by
unidirectional service flow in traditional mobile networks.
Coincidentally, the key NEs, such as HGU, OLT and BRAS, in fixed networks
also evolve with the control plane separated from the forwarding plane to provide
flexible on-demand network services for governments, enterprises, and households.
The control plane is deployed in a centralized manner, while the forward plane is
simplified and standardized, and managed in a unified way by SDN controllers through
NC/Yang, realizing flexible and efficient service orchestration. The CORD project, led
by AT&T and the Linux foundation, is a classic example. It is a major challenge in the
evolution of fixed networks how to further improve and optimize the system
architecture based on the existing achievements in the CORD project. And it has
become a big topic of fixed network evolution that puts forward a more practical,
efficient, fast, and intensive network evolution solution with a view to the service
demand for edge communication cloud and edge service cloud of wired and wireless
networks.
As shown in figure 2.2, with the evolution of the network architecture that the
control plane separated from the forwarding plane, traditional closed proprietary
equipment systems are gradually disintegrating, and transforming into cloud-based
open systems based on universal hardware and SDN/NFV open system. They carry
telecommunication network functions based on virtual machines and containers,
orchestrate services and resources in a unified way through MANO and cloud
management, and construct the integrated DevOps capability to significantly reduce the
time to come into market of the new services, implementing in-depth convergence of
IT and CT and enhancing the competitiveness of operators in services. Therefore, it is
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
6 Copyright©China Unicom Network Technology Research Institute, 2018
an inevitable choice to reconstruct telecommunication infrastructure based on cloud
technologies from central cloud to edge cloud.
Figure 2.2 Trend of Network Cloudification Evolution of China Unicom
As a product of ICT convergence, the edge computing–based platform can support
network function virtualization deployment, such as Central Unit, UPF, vSGW-
U/vPGW-U, vCPE, vBRAS, and vOLT. In addition, operators can make the storage,
computing, network, and security capabilities of the platform open up to third-party
application developers and content providers, and provide OTT applications with
unified edge deployment and management. Further, the edge computing platform can
abstract telecom network capability information into a variety of services (for example,
RNIS, location services, bandwidth management, and TCP optimization), and make
them open up to third-party applications and vertical industries to improve their service
performance and competitiveness among their peers.
The cloud-based edge computing platform allows operators to flexibly deploy
network functions and support infrastructure for fixed-mobile network convergence,
and the hosting and infrastructure service capabilities of the platform itself can help
third-party applications improve user experience, maximizing the value of both
applications and networks. This ensures that operators realize the digital transformation
from pipeline builders to eco-builders.
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
7 Copyright©China Unicom Network Technology Research Institute, 2018
3 Edge-Cloud Platform Architecture and Evolution
3.1 Overall Layout of Communication Cloud
Communication cloud is the network infrastructure that is constructed in a unified
way based on the existing network deployment and operation experience of operators
and uses SDN/NFV/cloud computing as the core technology to support the cloud-based
evolution of networks and match network transformation deployment. In the future, all
the NEs in 5G networks will be deployed based on cloud and NFV. To follow the trend,
China Unicom plans to transform traditional end offices into data centers based on the
Central Office Re-architected as a Data Center (CORD). The network architecture in
the future uses DC-centered 3-level communication cloud–based DC layout, and
deploys and builds edge, local and regional DCs at different levels to plan the
communication cloud resource pool in a unified way, achieving unified access bearers
and services for fixed networks, mobile networks, IoT, and enterprise dedicated lines.
In a cloud-based network, the lateral architecture still consists of the access networks
in traditional communication systems, MANs, and backbone networks, while the
vertical architecture is hierarchical, composed of transmission bearing, IP, NFV, and
network management and orchestration.
Figure 3.1 Cloud-Based Network Architecture of China Unicom Communication Cloud
As shown in Figure 3.1, the communication cloud–based network of China
Unicom is generally deployed at four layers, including three layers of DCs and one
layer of access CO. The features are as follows:
1) Regional DC: There are about 70 to 80 regional DCs across China with the
end-to-end delay less than 50 ms, which are used for the services and control plane NEs
(such as group OSS/NFVOs, provincial cloud pipe platforms, NFVOs, and VNFMs) of
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
8 Copyright©China Unicom Network Technology Research Institute, 2018
the whole country, large regions, and provinces. Regional DCs mainly serve for the
control NEs at the provincial level and the group level, including the IMS, GW-C, AMF,
SMF, MME, and NB-IoT core network NEs. Regional DCs are an important part of
communication cloud management. They mainly manage NEs, networks, and
infrastructure in a unified way. Regional DCs also realize the collaboration between
NFVOs and OSSs. An NFVO provides orchestration, resource management, and fault
management for virtual networks. An OSS coordinates services and resources, and
manages traditional networks. NFVOs work with OSSs for unified management of
traditional networks and virtual networks.
2) Local DC: There are about 600 to 700 local DCs across China with the end-
to-end delay less than 20 ms, which are deployed in prefecture-level cities and key
county-level cities of a province. Local DCs mainly serve for the control plane NEs in
MANs, centralized media plane NEs, services of local networks, control plane NEs,
and some user plane NEs, including the CDN, SBC, BGN-C, UPF, and GW-U NEs.
3) Edge DC: There are about 6,000 to 7,000 edge DCs across China with the
end-to-end delay less than 10 ms. The target of edge DC–based transformation is
converging equipment rooms. Edge DCs serve for access layer NEs and edge
computing NEs. They terminate and forward media streams. With the characteristics of
a low latency and high bandwidth, the future 5G RAN-CU, MEC, BNG-U, OLT-U,
and UPF NEs can be flexibly deployed at edge DCs, and provide location perception
and wireless network information for network edge users. Edge DCs allow cloud
service environment, computing, storage, network, and acceleration resources to be
deployed on network edges, realizing closer combination of applications and networks
and providing more network resources and services for users.
4) Access CO: There are about 60,000 to 70,000 access COs across China with
the end-to-end delay between 2 ms and 5 ms. Access COs achieve unified access and
bearing for the public, governments and enterprises, and users of fixed and mobile
networks to make resources more intensive and provide ultimate experience for users.
Access COs are widely distributed in villages and towns. The scenarios with a high
requirement for the delay and bandwidth, such as 5G CUs, DUs, MECs, and vCPEs,
can be deployed in access COs as required or directly deployed in the existing
equipment rooms. It is difficult to transform an access CO because an access CO mainly
accommodates access and traffic forwarding equipment and the condition in an access
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
9 Copyright©China Unicom Network Technology Research Institute, 2018
CO is worse than that in an edge DC. Therefore, the DC-based transformation of the
unified infrastructure for access COs are not considered at present (province-level
branch China Unicom can transform the access COs according to the actual service
needs).
3.2 Edge-Cloud Platform Architecture
The construction of the Edge-Cloud platform involves the infrastructure layer,
virtualization layer, edge computing platform layer, and upper-layer APPs. China
Unicom is committed to building an open and open-source edge services Edge-Cloud
platform, aiming to provide rich network capabilities, services and unified APIs for
developers, see Figure 3.2.
Figure 3.2 China Unicom’s Edge- Cloud Platform Architecture
3.2.1 Infrastructure Architecture
The business feature of the Edge-Cloud edge service platform determines that the
platform must be close to terminals and users in the network to provide the best
experience. Therefore, in edge service platform deployment, the positions of the access
CO and edge DC are key points requiring attention in the communication cloud
architecture.
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
10 Copyright©China Unicom Network Technology Research Institute, 2018
• The edge DC implements traffic convergence for the access network and leads
traffic to the local DC. The IDC can be a reference in guiding the DC reconstruction of
the infrastructure level, and similar environmental conditions of the IDC should be
implemented. The DC should have specified operating space, bearing capacity, power
supply capacity, and heat dissipation capability, and should support deployment of
general COTS devices. It is acceptable to use customized servers provided by device
manufacturers to satisfy requirements for many factors including the limited
deployment space and power supply.
• The access CO is relatively closer to terminals and users, aiming to increase
resource intensity and achieve perfect user experience. In most cases, access devices
and traffic forwarding devices are deployed in the access CO, and there is not an urgent
demand for DC reconstruction. The edge-service platform can be directly settled based
on the actual equipment room condition. In this case, general COTS devices are not
suitable for deployment due to the size, power consumption, and weight.
Figure 3.3 Server Evolution
In the IT field, the COTS server has evolved through several stages, including
tower server, rack server, blade server, and customized server, see Figure 3.3. The tower
server is mainly oriented towards scenarios with entry-level low-cost applications
scenarios. Based on the tower server, the rack server has higher computing density and
good extensibility through industrial standardization. The blade server not only has
higher computing density but also owns the advantages in costs, deployment, and
serviceability through sharing fans, power supplies, and networks among subracks and
providing built-in network solutions. With the development of cloud computing and
large-scale DCs, the above server types can no longer satisfy the operators' demands in
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
11 Copyright©China Unicom Network Technology Research Institute, 2018
costs, performance, and energy efficiency. Therefore, server customization, including
the multi-node form and integrated cabinet form, has become the focus of Internet
companies, aiming to reduce costs and share resources to the maximum extent.
In the CT field, with the development of the NFV technology, the traditional
private hardware platform of device manufacturers has also evolved towards the open
general COTS server. However, the general COTS server still has problems in several
aspects, including the supports of telecom feature, installation environment restrictions,
and capability of system architecture evolution, mainly reflected in:
• The design philosophy of rack servers may cause a waste of edge DC
equipment room resources. For example, for computing-intensive services (such as 5G
CU), the design of using more local hard disks and IO extensions may cause a waste of
space resources and costs.
• The blade server requires a great installation height and power consumption
over 5 kW, posing a considerable challenge in power supply, heat dissipation, and
bearing capacity. It may cause that the edge DC equipment rooms are difficult to
reconstructed or require huge costs.
• The depths of existing cabinets differ from the required cabinet depth. There
are a lot of 600–800 mm cabinets in the existing edge DC equipment rooms. However,
the general COTS devices should be at least 1 m deep.
• The bearing capacity requirement for equipment rooms is high. In most cases,
the bearing capacity of the IDC reaches 600 kg/m2, but the bearing capacity of edge
DC equipment rooms is far less than the value. Thus an additional constraint is added
on the deployment.
• I/O and resource acceleration are not designed in accordance with the NUMA
Balance technology, causing difference in performance and delay among cores.
• The management interfaces, protocols, and data formats on servers of different
vendors differ from each other. Remote fault locating, diagnosis, and troubleshooting
capabilities are not adequate, unable to satisfy the management requirements for
equipment rooms in unattended mode.
• The BIOS, FW, and OS configurations of servers of different vendors differ
from each other, causing the VNF unable to be migrated seamlessly and deployed
dynamically.
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
12 Copyright©China Unicom Network Technology Research Institute, 2018
• The specific –48 V power supply for the communication industry and the high-
voltage DC power supply possibly used in the future are not compatible with the AC
mainly used for general COTS devices.
Therefore, the edge-service platform has specific restrictions in consideration of
the hardware architecture, including the position, space, power supply, and service
scenario. It requires customized development and configurations. For example, the
following capabilities should be considered: data plane implementation and extension
(virtualized, unified and open API), rapid traffic offloading (intelligent network cards),
real-time codec capability (GPU), rapid image recognition (GPU/FPGA), and even
artificial intelligence. Both are important drivers of personalized openness and
configuration. Local computing, storage, I/O balancing, hardware acceleration, high
integration, and device Energy Efficiency Ratio (EER) are key factors that should be
considered in the future. Servers are customized designed in accordance with edge
application environments and edge service requirements to achieve the best solution.
The following factors should be considered in customization of general servers:
• General X86 architecture.
• Modular design to achieve flexible function extension on demand, for example,
hardware acceleration to satisfy service performance requirements.
• Flexible configurations in accordance with the requirements for computing,
storage, network, and hardware acceleration.
• Built-in network to provide better O&M experience through simplified
network management and topology architecture.
• Support for telecom-class features such as electromagnetic compatibility and
environmental adaptability. Considering various deployment positions, the edge server
must possess adaptability to extreme environments, for example, the server must be
able to operate in the 45°C environment for a certain period of time.
• Matching existing condition of equipment rooms and minimize equipment
room reconstruction.
• Proper physical specifications, including subrack height (3–4U. A great height
may cause heavy weight and large costs while a small height may reduce the gain
brought by common part sharing), subrack depth (800 mm cabinet installation
supported), and front maintenance mode (better O&M experience).
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
13 Copyright©China Unicom Network Technology Research Institute, 2018
Behind the most general concept of COTS device, the real core requirement of
customers is decoupling of software from hardware. It means that a customer can
flexibly deploy services without being restricted by hardware device manufacturers.
From this perspective, the customized general servers based on the X86 architecture
can be classified to the range of COTS devices. In addition to the features of decoupling
software from hardware, virtualization, and flexible service deployment, customization
of general servers also involves scenario-based device appearance and device
architecture, aiming to achieve higher energy efficiency and better cost-effectiveness.
Therefore, the customized general server is comparatively the best hardware choice for
the edge DC and access CO.
3.2.2 Virtualized Layer Architecture
VIM refers to the virtual resource management platform which manages the
NFVIs in a domain. In the hierarchical DC architecture, each edge DC, local DC, and
regional DC are all respectively corresponds to an NFVI, and all the NFVIs are
managed by the VIM. China Unicom recommends to use that the OpenStack, enhanced
VIM based on open-source Openstack, or VIM based on the Kubernetes container
technology.
Compared with regional DCs and local DCs, edge DCs have different scales and
carry various types of services, and thus have flexible deployment modes.
Reconstruction of an edge DC should be implemented in stages. In the initial stage,
virtualized reconstruction can be implemented at some hot spots with good deployment
conditions. For large-scale edge DCs, each edge DC is deployed with Openstack. For
small-scale edge DCs, deploying Openstack for each edge DC may cause a huge waste
of resources. Thus, it is recommended that compact Openstack deployment or
distributed OpenStack deployment can be used locally. In local compact Openstack
deployment mode, resources can be saved by deploying the control node and computing
node as one node and decouples some components to the actual requirements. In
distributed Openstack deployment mode, only computing modes are deployed in edge
DCs with fewer nodes, control nodes are deployed in regional DCs or edge DCs with
more nodes, and all edge DCs are managed in a unified manner.
The container technology provides better elasticity response speed, system
capacity flexibility, and computing resource utilization. Considering the development
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
14 Copyright©China Unicom Network Technology Research Institute, 2018
of new services and technology evolution trend, the virtualization layer should support
the container technology to construct distributed container clouds. The container
technology supports two deployment modes: container deploys in a physical machine
and container deploys in a virtual machine. In container deployment scenarios, the
Cloud Container Engine should be deployed on the VIM to provide highly reliable and
high-performance management services for enterprise container applications, support
native applications and tools of the Kubernetes community, and simplify the
construction of automation container operating environment on the cloud.
The low delay and high-speed forwarding features of the edge computing service
require that the virtualization layer should have the hardware acceleration capability.
The VIM provides unified interfaces to adapt to various accelerator types. Accelerators
are abstracted and managed in a virtualized manner as computing devices, storage
devices, and networks. Accelerators are presented as acceleration resources and provide
comprehensive acceleration services. With the development of the NFV standard,
hardware acceleration has become increasingly important. The standardization of
functional interfaces of accelerator resources is being gradually conducted. Openstack
has started the Cyborg project, aiming to provide a general hardware acceleration
management framework. The acceleration hardware includes the encryption card,
intelligent network adapter, GPU, and FPGA. China Unicom is actively promoting
relevant research.
Figure 3.4 Edge Cloud Management Platform
In ETSI NFV MANO specification, the VIM executes the virtual resource
management only. However, the edge cloud requires an integrated platform having the
VIM, PIM, and service management capabilities, which is known as the Cloud
Management Platform (CMP). For edge clouds, the CMP should support hierarchical
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
15 Copyright©China Unicom Network Technology Research Institute, 2018
architecture, for example, the edge cloud management platform distributed in provinces
and cities and the unified cloud management platform of the group. As shown in Figure
3.4, an edge DC and its access COs comprise an edge cloud, and an edge cloud
management platform manages one or multiple edge clouds. The edge cloud
management platform can be deployed either in the edge DC or in the regional DC in
accordance with the actual condition. It is logically interconnected with multiple VIMs,
and provides access point convergence, unified management of heterogeneous
resources, and position-based resource scheduling based on VIM virtualized
management.
Figure 3.5 Edge Cloud Management Platform Functional Diagram
As shown in Figure 3.5, the edge cloud management platform provides functions
including cloud resource management, O&M management, edge cloud service
management, and capability openness. The cloud resource management connector
(adaptation layer) accesses the VIM (OpenStack, K8S) of the edge DC through an plug-
in or agent to form a geographically distributed but logically unified cloud resource
pool (computing resources, network resources, storage resources, and acceleration
resources) to implement unified management and scheduling of heterogeneous
resources. The O&M module implements unified monitoring, alarm management,
performance management, and log management of resources. At the early stage of edge
cloud computing construction, the ecological chain has not been formed. To implement
rapid validation of new services and put the services into operation quickly, the NEs
(for example, MEPO, MEPM, and VNFM) in the MANO framework coordinating with
NFVO to implement VNF and third-party vApp lifecycle management can be packaged
and delivered in a unified manner as the components of the edge cloud management
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
16 Copyright©China Unicom Network Technology Research Institute, 2018
platform. After the edge cloud ecological chain becomes mature, these components can
be split off from the edge cloud management platform and be integrated into the MANO
framework. The edge cloud management platform is interconnected with the unified
cloud management platform and NFVO through the unified capability openness layer.
The unified portal provides a unified GUI for cloud resource presentation and operation.
It is used for unified management of edge cloud resources and provisioning and
maintenance of local edge cloud services.
3.2.3 Service Orchestration and Management Architecture
In the overall communication cloud layout of China Unicom, regional MANO and
the unified cloud management platform are deployed for regional DCs to implement
communication cloud infrastructure management. MANO management includes:
• NFVO: Based on the authorization of the unified cloud management platform,
this module manages and orchestrates authorized resources in a unified manner,
including definitions, collaborative scheduling, and lifecycle management of resources
and services, to put the services into operation quickly. Cross-regional services are
orchestrated by the group NFVO.
• VNFM: This module is responsible for lifecycle management for VNF NEs,
including creation, modification, deletion, and elasticity expansion of VNF NEs.
The traditional MANO can be used to manage ETSI-NFV VNFs only. On the
Edge-Cloud service platform, not only VNFs but also third-party edge applications
(Edge-APPs) are deployed. Thus, Edge-APPs should also be orchestrated and managed.
The Edge-APP requirements fall into two categories:
1) Application package requirement
• The following information should be included: software image (or image link),
image format (for example, virtual machine, container), application description
(requirements and rules which applications depend on or prefer to), provider signature
(including the abstract and public key of the signature and relevant certificates).
• The files in an application package can be signed individually. For each signed
file, the corresponding public key, algorithm, and certificate are stored in a position
known to all in the application package.
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
17 Copyright©China Unicom Network Technology Research Institute, 2018
2) Application description requirement
• Includes the minimum computing resources (for example, quantity, features,
and functions of virtual machines), virtual storage resources, and virtual network
resources required by the application.
• Includes the (additional) service list and (additional) feature list required for
the application operation.
• Supports descriptions of traffic rules, DNS rules, and delays descriptions for
Edge-APPs.
The requirements for computing, storage, and network resources are consistent
with the VNF requirements, and can be parsed and processed through traditional
MANO. However, the requirements for description of delays, traffic rules, and DNS
rules cannot be parsed and processed through traditional MANO. Considering the
special requirements for third-party Edge-APPs, China Unicom extends the traditional
MANO management domain, finally realizes the management of the edge service
platforms and Edge-APPs.
As shown in Figure 3.6, based on NFVO and VNFM, two logical function modules
are added: MEP-O and MEP-M.
• MEP-O: This module collaborates with NFVO to implement version package
management (including loading, querying, enabling, disabling, and deleting version
packages) and application lifecycle management. MEP-O supports parsing and directly
processing descriptions related to Edge-APPs, for example, it can select a proper edge
service platform to configure features for the MEP-M. For common descriptions, MEP-
O interacts with NFVO and uses NFVO to orchestrate resources, and orchestrates NFV
network services for a group of Edge-APPs.
• MEP-M: This module manages NEs on the edge service platform and Edge-
APP rules and requirements (for example, traffic rules and DNS rules), and implements
lifecycle management for the edge service platform and Edge-APPs.
The unified Cloud Management Platform (CMP) manages infrastructure layers of
multiple DCs within its management areas, divides the resource pool and manages
authorities, and monitors various types of information of virtualized resources and non-
virtualized resources, including topology, alarms, performance, and capacity. It also
provides reports related to whole resources, and helps O&M personnel evaluate
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
18 Copyright©China Unicom Network Technology Research Institute, 2018
potential risks based on AI algorithms to avoid or eliminate problems in time.
Considering that Edge-Cloud is mainly oriented to vertical industries, CMP can provide
service portals oriented to third parties to support MEP capability openness , Edge-APP
lifecycle management and Devops, promoting the construction of edge content ecology.
Figure 3.6 Edge-Cloud Platform and MANO Architecture of China Unicom
As defined in the ETSI GS IFA 009, NFVO has two functional modules. The RO
module is responsible for cross-VIM NFVI resource orchestration while the NSO
module is responsible for NS LCM. From the perspective of long-term ICT evolution,
CMP, RO of NFVO, and VNFM can be integrated into a functional entity to provide a
unified Automation & Orchestration platform for IT and CT applications.
3.2.4 Edge-Cloud Platform Capabilities
China Unicom's Edge-Cloud platform not only provides various basic services but
also provides a unified capability openness framework by formulating MEC-Enabled
API specifications on the basis of ETSI standards. This platform serves and manages
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
19 Copyright©China Unicom Network Technology Research Institute, 2018
third-party applications. Figure 3.7 shows the capability openness framework of China
Unicom Edge-Cloud platform. The service capabilities of basic networks can be
provided to third-party applications in a safe and efficient manner, and can be reliably
shared between third-party applications. This promotes rich edge application ecology
and meets various edge application scenarios.
Figure 3.7 Capability Openness Framework of the Edge-Cloud platform
3.2.4.1 Edge-APP Management Capability
The APPs deployed on the Edge-Cloud platform can obtain necessary services,
and can also provide services for other users. To achieve the above purposes, the Edge-
Cloud platform needs to support APP management. The ESTI-MEC specification
defines the interface between the Edge-Cloud platform and APPs as the Mp1 interface.
Based on the Mp1 interface specified by ETSI-MEC, China Unicom defines a unified
APP management framework, which provides the following functions for third-party
applications:
• Service registration and discovery
• Service termination
• Subscription and notification of service-related events
• Service API and access control
• Authentication and authorization of service APIs
• Statistics and logs of service APIs
• Flow control rule configuration
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
20 Copyright©China Unicom Network Technology Research Institute, 2018
• DNS rule configuration
A third-party application can obtain the capabilities provided by the platform and
provide the capabilities for other applications through the platform. The descriptions of
specific capabilities and related service APIs should be provided by the third-party
application developer.
3.2.4.2 Location Service Capability
User location information is a key factor to associate online services with real
scenarios, especially in large-sized super markets, railway stations, and airports with
dense population and high traffic, where the requirements of accurate sales, online
payment, mobile advertising, indoor location service, operation innovation, and service
innovation are growing fast. The Edge-Cloud based indoor location solution uses the
fingerprint database matching location algorithm based on uplink SRS measurement
results. The location procedure is divided into two phases:
• Offline fingerprint database collection: collects the SRS fingerprint data in the
location area and stores it in the Edge-Cloud. The data includes the coordinates of each
indoor position and the signal strength array, which are one-to-one matched.
• Online location: Queries the fingerprint database to find K similar samples in
accordance with the real-time RRU measurement data of each UE to be located, and
calculates the real-time position coordinates of the UE in accordance with the k-Nearest
Neighbor (KNN) algorithm.
An indoor eNodeB reports the UE position coordinates and UE ID (for example,
UE IP) through the API defined by the Edge-Cloud platform. This platform can either
provide the user location service directly or provide a customized interactive service by
using the real-time IDs, services, and behaviors of authorized UEs in the mobile
network.
3.2.4.3 Traffic Breakout Capability
The breakout capability based on the Edge-Cloud platform can shunt some
services to the local network, for example, a campus network, or to the applications
deployed on the edge service platform to reduce the pressure on transmission bandwidth,
decrease delay, and improve user experience of the low-latency and high-bandwidth
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
21 Copyright©China Unicom Network Technology Research Institute, 2018
services. For the entire architecture of local breakout, see Figure 3.8. The procedure is
described as follows:
• In the uplink, the Edge-Cloud platform transparently transports the data that
does not need to be processed in the local network to the mobile core network, and
decapsulates the GTP headers of the data that needs to be processed and shunt packets
to the local network.
• In the downlink, the Edge-Cloud platform transparently transports the data that
is not generated by a non-local network, for example, generated by Internet, to eNodeBs.
Similarly encapsulates the data that is generated by the local network into S1-U packets,
and sends them to eNodeBs.
Local breakout rules include breakout by domain name, address, and user ID.
Meanwhile, the network will be a converged network containing 4G, 5G, and WiFi, and
the Edge-Cloud platform needs to support network convergence breakout.
Figure 3.8 Diagram of Service Breakout Based on the Edge-Cloud Platform
3.2.4.4 Real-Time Codec Service Capability
To support media-level services, for example, local live show, local video
surveillance, and video cloud games, the edge service platform needs to support the
real-time codec service capability.
1) Live show: The camera collects audio and video signals and pushes them to
the media server deployed on the edge service platform. To adapt to more video playing
terminals, the media server needs to support real-time transcoding and video frame re-
encapsulation.
2) VR games: To provide perfect and shocked experience, VR games need CPUs
and GPUs with a high computing rate and large computing cost, which prevent VR
games from being popular. If VR cloud games are deployed on network edge through
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
22 Copyright©China Unicom Network Technology Research Institute, 2018
the Edge-Cloud platform, the strong computing capability on pictures and real-time
media of the Edge-Cloud platform can accelerate VR game popularization.
3) Video communication: In a video call, especially a multi-party call, each data
stream needs to be coded, decoded, and synthesized on the UE and cloud. Deploying
the codec capability on the edge service platform can solve both the codec
synchronization problem caused by inconsistent UE capabilities and the bottleneck
problem caused by centralized coding, decoding, and synthesis. The distributed video
communication coding and decoding capability based on the edge service platform,
combined with the UE capability and cloud capability, can allocate network resources
in a reasonable manner and provide better user experience.
3.2.4.5 Intelligent Analysis Service Capability
Now the video surveillance service becomes one of the major services of Smart
City. Video surveillance generates lots of backhaul traffic, while most pictures are still
or worthless. Now the video surveillance service uploads all video streams to the back-
end server for processing, with high cost and low efficiency. Deploying edge
intelligence capability on the edge service platform to implement intelligent video
analysis can either return the changed pictures only or send structural data processed
by intelligent analysis to the DC, saving the transmission resources of the backhaul
network. To avoid high latency, bad experience, and high backhaul bandwidth cost due
to lots of alternative service routes, the low-value surveillance contents can be saved on
the local server. In addition, distributed video storage based on the edge service
platform can reduce the bottleneck problem caused by centralized storage.
3.2.4.6 Radio Network Information Service Capability
Deployed on network edge, the Edge-Cloud platform facilitates the awareness of
real-time radio network information. The Radio Network Information Service (RNIS)
includes the real-time radio network conditions, measurement and statistics information
on the user plane, for example, CQI, SINR, and BLER, and UE bearer information, for
example, UE context and RAB.
The RNIS provides open interfaces in the format of API to third-party applications
through the Edge-Cloud platform, helping the applications optimize service procedures,
improving user experience, and implementing deep convergence of the network and
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
23 Copyright©China Unicom Network Technology Research Institute, 2018
services. Based on the radio network information awareness capability provided by the
Edge-Cloud platform, third-party applications can obtain differentiated network
services in accordance with their requirements, at last increasing the customer
satisfaction level.
3.2.4.7 User Identification Service Capability
The edge service platform authenticates third-party applications, and accepts the
API requests from only the authorized third-party applications and forwards the
requests to the internal services of the Edge-Cloud platform. Authorized Edge-Cloud
application instances can activate or deactivate the associated configuration rules by
user ID. By means of using the identification service, third-party applications can map
external application IDs to the internal IDs in the mobile network to implement data
operations for specific users.
3.2.4.8 Bandwidth Management Service Capability
After being deployed on the edge computation platform, various applications
require different network resources, for example, the bandwidth. The bandwidth
management service can set bandwidth parameters on different applications, for
example, bandwidth, priority, and allocation policy of static or dynamic bandwidth, and
manage the parameters in accordance with the service policies of the third-party
applications in a unified management.
3.2.4.9 QoS Service Capability
With the increase of video-type services, QoS on the radio access network side
will be increasingly focused. And more and more users, including individual users and
enterprise users, want to obtain different user experience. Research on how to schedule
air-interface resources for HD video services becomes a new direction. In the high-user-
density spots such as airports, railway stations, and campuses, the Edge-Cloud platform
can provide different QoS services for users acquiring the same type of services by
detecting real-time data, for example, the user requirement class, service type, and
network situation. For the video QoS optimization diagram, see Figure 3.9.
The Edge-Cloud platform can identify the service type and user type in accordance
with user data packets, and provides QoS information for eNodeBs in accordance with
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
24 Copyright©China Unicom Network Technology Research Institute, 2018
the analysis results. eNodeBs provide different radio bandwidth and latency guarantee
for the data packets of different users. By integrating OTT service information and radio
access network information, the Edge-Cloud platform can use the advantages of
intelligent channels to guarantee QoS of key users and services.
Figure 3.9 Video QoS Optimization Diagram
3.2.4.10 Procedure Collection and Charging Service Capability
In traditional networks, the core network NEs implement charging function. Due
to the location change of the Edge-Cloud platform and deployment of third-party
applications, the charging service has new requirements.
• Charging requirements related to APP providers: charging on the operation
procedure of third-party applications in accordance with the factors related to third-
party applications, for example, the number of times that an APP service is used and
service duration. Third-party applications reports the generated charging information to
the third-party charging center to which they belong.
• Charging requirements related to operators: charging on third-party
applications in accordance with statistics information, for example, the number of times
that an APP is on boarding, number of times that an APP is instantiated, resource
consumption data, number of times that an APP is migrated, number of times that a
service is used, and service duration. The charging data generated by the Edge-Cloud
should be sent to the charging center of the operator.
On account of 5G specifications are not frozen yet, the charging capability will be
developed in multiple ways, for example, the specification, technology, and engineering.
The detailed solution is described as follows:
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
25 Copyright©China Unicom Network Technology Research Institute, 2018
The UPF implements charging on users. The UPF forwards 5G user data to the
Edge-Cloud platform, implements charging based on the charging policy, guarantees
security, and reports charging information to the charging center. Accurate charging on
users is supported either before or after the Edge-Cloud is deployed.
The Edge-Cloud platform implements charging on APPs/APIs. The Edge-Cloud
needs to monitor the events that can be charged, generates charging records in
accordance with the events, and reports the records to the charging gateway. The events
that can be charged are classified into three types:
• Resource occupation events: provide the number of occupied virtualized
resources and/or the usage duration.
• Management and orchestration events: The Edge-Cloud management elements
invoke the API to implement APP management functions, for example, creating or
deleting APPs, and migrating APPs.
• Capability openness API invoking events: APPs invoke the API and use the
services provided by the Edge-Cloud platform or network.
In addition, the Edge-Cloud platform needs to support online charging and offline
charging.
• Offline charging: The Edge-Cloud periodically collects and reports the
resource occupation data and management/orchestration data to the offline charging
function module of the network. The charging system generates charging data of
consumers when a bill period ends.
• Online charging: The Edge-Cloud monitors resource occupation. When the
quota is being used up, the Edge-Cloud reports resource occupation to the online
charging function module and requests a new quota. If your account balance is
insufficient, the Edge-Cloud rejects the quota request, and suspends the resources being
used.
3.2.4.11 IoT Service Capability
If all IoT data is processed on the cloud, network traffic and application
instantaneity become new challenges. For the applications that require extra-high
instantaneity, for example, detection, control, and execution applications for industry
systems, and the multimedia applications that generate massive data, the analysis and
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
26 Copyright©China Unicom Network Technology Research Institute, 2018
decision functions of some data can be deployed on network edge. For the applications
that require high privacy, the IoT capability can be moved to the network edge or
campus to satisfy the security requirements of specific users.
Performing data analysis and pre-processing on the Edge-Cloud can efficiently
deal with data explosion and reduce network traffic. The Edge-Cloud can decrease the
device response time and reduce data from devices to cloud data centers to efficiently
allocate network resources. The Edge-Cloud can be deployed inside a campus to ensure
data privacy. The Edge-Cloud platform supports the following IoT capabilities:
• Multi-access capability: allows UEs and gateways to access the network
through multiple protocols, including TCP, UDP, MQTT, and COAP, and supports
multiple types of UEs, for example, LORA, ZIGGEE, and NB-IoT.
• Device management: supports device status management, map and location
management, and storage, display, and distribution of the data reported by devices.
• Data analysis: supports analysis and pre-processing of the data reported by
devices.
• Rule management: supports operations on the data reported by devices in
accordance with the user-defined rules, for example, generating alarms, and operating
the device or other related devices.
• Coordination management: supports coordination with the higher-layer cloud
platform, and implements convergence, storage, and long-term analysis of important
data on high-layer cloud platform nodes.
With the service development and network evolution, the Edge-Cloud platform,
based on the capabilities mentioned above, will bear more capabilities and applications
to accelerate fixed-mobile convergence.
3.3 Evolution Roadmap of the Edge-Cloud Platform
Figure 3.10 shows the evolution planning of China Unicom’s Edge-Cloud
platform.
1) Large-scale pilot projects (September 2017-June 2018)
China Unicom began large-scale pilot projects from September 2017. By the end
of June 2018, such projects will be constructed in China's 15 provinces: Beijing, Tianjin,
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
27 Copyright©China Unicom Network Technology Research Institute, 2018
Shanghai, Sichuan, Zhejiang, Chongqing, Guangdong, Henan, Liaoning, Fujian,
Jiangsu, Hubei, Hebei, Hunan, and Shandong. The application scenarios include smart
campuses, smart venues, smart parks, industrial Internet, and IoV.
During this stage, the Edge-Cloud platform can implement software&hardware
decoupling (COTS&Cloud OS decoupling). The Cloud OS and MEP manufactured by
the same vendor are deployed. It is recommended that the Edge-APP and MEP can be
deployed onto the same Cloud OS and the MEP Mp1 rule be followed. The ETSI MEC
ISG has not strictly defined API standards, and there are great differences among
platforms manufactured by different vendors. It is not needed to unify the standards of
northbound API interfaces between the Edge-APP (manufactured by the same vendor
as the MEP) and MEP. China Unicom has begun the formulation of unified northbound
API standards that will be completed in June 2018.
Figure 3.10 Evolution Planning of China Unicom's Edge-Cloud Platform
2) On-demand commercial deployment of 4G (from July 2018 to December
2018)
In accordance with the early-stage Edge-Cloud pilot network construction and
service demonstration effect, China Unicom will replicate and promote the success
stories of the incubated edge services and implement commercial construction step by
step. During this stage, the Edge-Cloud platform will implement 4-level decoupling
between the COTS, Cloud OS, MEP, and Edge-APP, and embed the VCPE and BNG-
U functions by using the same COTS and different Cloud OSs.
On the Edge-Cloud platform, third-party application developers develop the Edge-
APP in accordance with China Unicom's unified API specifications, and are not
restricted by the platforms manufactured by different vendors, thus greatly shortening
the service launching low. In addition, China Unicom will begin the preparation of the
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
28 Copyright©China Unicom Network Technology Research Institute, 2018
equipment rooms of edge DCs in multiple key provinces (including 15 provinces where
pilot projects are located).
3) Pre-commercial construction of 5G networks (2019-2020)
China is actively promoting 5G research and industrial development, and raises
the goal of “Become a Pioneer in 5G”. With the rich Edge-Cloud pilot project and
commercial construction experience accumulated in early stages, China Unicom will
start the large-scale construction of edge-DC cloud resource pools in 2019 to boost the
pre-commercial use of 5G networks.
During a short period, the 5G CU, UPF, VCPE, and BNG-U manufactured by the
same vendors are deployed in the same Cloud OS, smoothly implementing
VNF&Cloud OS decoupling. The northbound API capabilities between the MEP and
Edge-APP will become richer (4-level decoupling is still kept). Considering the
differences between the CT domain and IT domain, the deployment across different
Cloud OSs is allowable. Considering the scenario where 4G and 5G networks will
coexist for a long time, the Edge-Cloud platform supports the simultaneous connection
of 4G and 5G base stations and traffic unloading.
4) Large-scale commercial use of 5G networks (2021-2025)
During this stage, 5G networks will be rapidly constructed and put into large-scale
commercial use. At the same time, the cloudification architecture of China Unicom’s
communication networks will be shaped. It is estimated that full cloudification
deployment will be achieved in 2025. China Unicom's Edge-Cloud platform supports
the deployment of the CU, UPF, GW-U, VCPE, BNG-U, MEC, and Edge-APP
manufactured by different vendors, completely implementing 4-level decoupling. The
Edge Cloud-oriented various services are more prosperous, and business modes are
more mature. China Unicom will build more sophisticated edge ecology and implement
service layout with the characteristics of a standard leader, full commercial use, self-
control, openness and open source.
China Unicom will gradually implement the seamless integration of the
business/enterprise private Ethernet, FTTP access network for family users, and FTTC
access network for mobile users in accordance with actual service requirements. The
vision is to provide a unified Edge-Cloud platform for FMC users.
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
29 Copyright©China Unicom Network Technology Research Institute, 2018
4 Edge-Cloud Standardization and Industrial Chain
4.1 Standardization Progress
In 2014, the ETSI first initiated the MEC standard project that aimed to build a
service platform with cloud-based computing and an IT-based environment at the edge
of mobile networks for application developers and content providers, and to open
wireless network information through the platform, supporting and locally managing
high-bandwidth and low-latency services. Some leading vendors and operators, such as
Vodafone, ZTE, Intel, Huawei, Nokia, and HP are members of the alliance. Phase I of
the project is completed, which is deployed based on traditional 4G network
architecture and defines the service scenarios, MEC system reference architecture,
MEC platform application enablement APIs, framework of the system management and
operation, application lifecycle and management and wireless service capability APIs
(e.g. RNIS, location, and bandwidth management). To satisfy the requirements for
commercial MEC deployment and fixed-mobile convergence, ongoing Phase II of the
project focuses on multi-access edge computing systems, including 5G, WiFi, and fixed
networks, mainly involving the MEC in NFV reference architecture, end-to-end MEC
mobility, network slicing support, legal interception, container-based application
deployment, V2X support, and WiFi and fixed-network capability exposure.
With the extension of ETSI MEC influence, the 3GPP is also engaged in the
support research on edge computing. The 4G CUPS and 5G New Core introduce the
separation of the control plane from the forwarding plane. The forwarding plane is
deployed on the edge of wireless networks in a distributed way, and the control plane
deploys and controls the forwarding plane in a centralized manner, ensuring the on-
demand local distribution of services. The SA2 5G system architecture fully supports
edge computing in terms of local routing and service manipulation, session and service
continuity, network capability exposure, QoS and Policy Control. In addition, the
research on SA5 network functions management of next generation network
architecture and features, SA6 Common API Framework for 3GPP Northbound APIs
will further consider the demand for edge computing. The 3GPP is accelerating the
process of commercial edge computing as an effective supplement to ETSI MEC.
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
30 Copyright©China Unicom Network Technology Research Institute, 2018
Following the fixed-mobile convergence trend, the Broadband Forum is
converging wired and wireless networks by using the unified core network and network
slicing technology. The definition of the unified interface between an access network
and a core network enables 5G New Core to support the convergence of wired and
wireless services. This further extends the enabling capability of the multi-access edge
computing platform and support edge computing services for fixed-mobile convergence.
In addition to the efforts of the international organizations, China also makes
contributions. In 2017, the China Communications Standards Association (CCSA)
launched an edge computing research project. As a part of the project, China Unicom
initiated and led the research on the capability openness technology of the 5G edge
computing platform, which studies on capability exposure for edge computing in 5G.
The solutions will be based on the architecture of the edge computing platform and the
capability of mobile networks to standardize the framework and content for network
capability exposure. In January 2018, an international standard project, “IoT
Requirements for Edge Computing”, led by China Unicom, was successfully initiated
at the plenary meeting of ITU-T SG20 WP1, which was the first time that an edge
computing project is initiated by the ITU-T in the field of industrial Internet. China
Unicom will lead the international standardization efforts of ITU-T IoT edge computing
to promote health and sustainable development of the edge computing industry.
However, edge computing is an ecosystem that includes applications, platforms,
and networks, so standard organizations cannot promote the rapid development of the
industry alone. Therefore, it is necessary for an edge computing industry alliance, for
example, OpenFog Consortium, to fully explore industry service scenarios, integrate
industry resources through operators, and combine application requirements with edge
computing platform standards, to promote the development of the edge computing
industry together.
4.2 Situation of the Industrial Chain
As shown in Figure 4.1, the Edge-Cloud industry is an ecosystem built by telecom
operators, telecom vendors, IT vendors, third-party APP developers, and content
providers. Telecom operators are the core of the whole industrial ecological chain, and
are also the foundation of industrial collaboration.
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
31 Copyright©China Unicom Network Technology Research Institute, 2018
Figure 4.1 Landscape of the Edge-Cloud Industrial Chain
It is predicted that the number of connected IoT devices will reach 50 billion in
2020, and 50 percent of these devices should have the edge computing capability. For
the global 5G market, it is estimated that the scale of Edge-Cloud devices and system
markets will reach $80 billion in 2021. In the whole Edge-Cloud industrial chain, the
ratio of pipeline connection value is only 10 to 15 percent, and the ratio of application
services is 45 to 65 percent. Telecom operators are successively beginning network
reconstruction and transformation and changing from traditional pipeline connectors to
industrial integrators, aiming to become service providers. By relying on the
establishment of edge cloud platforms, telecom operators can provide its platform
capabilities to third-party APP developers, rapidly launch user-oriented innovative
services, and shorten the Time to Market (TTM).
Figure 4.2 China Unicom's Edge Cloud Ecosystem
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
32 Copyright©China Unicom Network Technology Research Institute, 2018
China Unicom is committed to creating a brand-new value chain and an edge
ecosystem full of vitality by collaborating with all industries. The subjects of all
processes are motivated to work closely together to promote the integration among IT,
CT, and OT and deeply excavate the potential value of the edge network. As shown in
Figure 4.2, China Unicom focuses on big video, VR/AR, industrial IoT, IoV, and other
high-bandwidth and low-latency services by adhering to the “Openness and
Cooperation” philosophy. It provides location, charging, QoS, and other platform
capabilities to third parties, aiming to construct an Edge-Cloud content ecosystem with
industrial partners. Currently, China Unicom has over 100 Edge-Cloud partners,
including Tencent, Alibaba, Baidu, Jingdong, iQIYI, Intel, ZTE, Star Technology, H3C,
HPE and so on.
4.3 Orientation of China Unicom's Edge Cloud
Global operators are encountered with the dilemma in which voice services create
less value. At the same time, the rapid growth of data traffic does not bring proportional
revenue increase due to the coming of innovative enterprises. How to resolve the
dilemma and expand value-added services are the problems to be solved. During the
ICT integration tide, operators regard thousands of edge DCs as the best high-quality
resources compared with OTT. China Unicom will deeply excavate the Edge-Cloud
service platform capability, the same time, still enhancing pipeline capabilities, and
open a new window for OTT collaboration.
• A unified Edge-Cloud resource pool will be rapidly constructed during mixed-
ownership reforms, on the basis of 5G cloudification evolution, aiming at digital
transformation.
• An open, open-source, and friendly edge service PaaS platform is built to
provide rich platform capabilities and unified APIs for third-party APP developers.
• 4G networks are 5G-oriented and important service scenarios are deployed to
provide business support and incubation.
• The ecological collaboration depth and width of the Edge-Cloud industry will
be expanded, and all industrial partners are involved to explore business models and
co-build the edge ecology of 5G networks.
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
33 Copyright©China Unicom Network Technology Research Institute, 2018
5 Key Service Scenarios and Commerial Applications
5.1 Scenario 1: Private Customization of Campus Area Network
Campus Area Network generally refers to the private intranet of university or
enterprise, which covers key requirements, such as regular mobile office work and
internal enterprise communication, and needs to be isolated from public networks for
security. With the development of IoT, in addition to the demand for interpersonal
communications, that the demand for management of things is increasingly urgent.
WiFi-based dedicated mobile networks show the disadvantages of a long construction
period, high costs, high maintenance costs, and cannot ensure sufficient, stable, or
secure coverage or provide continuous services. For user experience, it is inconvenient
for end users that dedicated networks can be used only for special purposes. Public
services and private services are separated from each other is also inconvenient.
Figure 5.1 Customized Campus Area Network Solution
As shown in Figure 5.1, through the local breakout function, the Edge-Cloud
Service Platform can create a mobile virtual private network to shunt enterprise or
campus users’ local data services into edge servers, which forms a local 4G virtual
private network, ensuring the security and stability of mobile office work, avoiding
traffic alternation, and reducing the delay in service access. It can also help the
enterprise or campus create a private IoT network and deploy related services.
China Unicom is committed to building Edge-Cloud based “Smart Park Solutions”
in several provinces in China to shunt service access by enterprise or campus user and
provide carrier-class security identity authentication. Besides, the Edge-Cloud based
IoT management platform can manage wireless sensors such as for fire sprinklers and
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
34 Copyright©China Unicom Network Technology Research Institute, 2018
fire alarms jointly, perform indoor positioning–based rescue, and organize real-time fire
information–based evacuation to build a safe and reliable private network in a Smart
Park.
5.2 Scenario 2: Mobile vCDN
The rapid growth of video services significantly affects the bearer capacity of
mobile operators. OTT vendors have deployed many CDN nodes on a large scale.
However, the CDN nodes are mainly deployed inside fixed networks, and mobile users
can access video services only on the back end of the core network, which poses a huge
challenge to operators’ network resources, especially the transmission bandwidth on the
S1 interface. When CDN nodes are deployed inside a mobile network, for example,
vCDN nodes are migrated downward to the edge DCs of operators through the
deployment on the Edge-Cloud platform, the traffic load on traditional networks can be
greatly reduced and mobile user experience of video services can be improved.
Figure 5.2 Domain Name–Based Breakout and NAPT Function Procedure
Working with ZTE and Intel, China Unicom verified edge vCDN of Tencent
videos and Wo videos. Edge-Cloud shunts video services by domain name through the
NAPT function. As shown in Figure 5.2, the CDN HTTP DNS servers of OTT vendors
establish server mapping by IP address to ensure that the requests converted by Edge-
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
35 Copyright©China Unicom Network Technology Research Institute, 2018
Cloud NAPT can be mapped to local edge servers (edge vCDNs) according to the load
balancing strategy.
• The Edge-Cloud platform configures the specific domain name, listens to the
DNS response sent to the UE, obtains the IP address corresponding to the domain name
from the DNS response, and dynamically generates a distribution rule for the IP address.
• The Edge-Cloud platform matches the distribution rule, and performs NAPT
(the source IP address is replaced with a specific IP address) for the packet sent to the
HTTP DNS.
• The Edge-Cloud platform makes a reverse NAPT conversion on the packet
returned to the UE by the HTTP DNS and then sends it to the UE.
• The Edge-Cloud server establishes a breakout mechanism, and locally shunts
the packets whose destination address is the edge server.
As shown in figure 5.3, compared to provincial node server, the RTT latency
reduces 50% and HTTP download speed increases 43%. At the same time, China
Unicom together with ZTE have verified video optimization based on wireless
information in the existing networks. In the preliminary stage, the indicators about
video quality and wireless network quality are reported through the Wo video app.
Based on the reported indicators, the Edge-Cloud platform determines the adaptation
of the video bit rate and informs the UE to adjust the bit rate. The UE app reports a
video bit rate switchover instruction to the Wo video server and the server switches to
the corresponding video source.
Figure 5.3 Test result of vCDN deployed in Edge DC
5.3 Scenario 3: Video Monitoring and AI Analysis
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
36 Copyright©China Unicom Network Technology Research Institute, 2018
Traditional monitoring are deployed by using cables and all the cameras must be
connected to the cable network. So a large number of cables are used, resulting in high
costs but low efficiency. The monitoring videos are transmitted to the cloud or servers
for storage and handling through the bearer networks and core networks. This method
increases network burden and the end-to-end delay of services. To solve the problems,
monitoring data can be shunted to the Edge-Cloud service platform, effectively
reducing the network transmission load and the end-to-end service delay. Video
monitoring can also be combined with AI. The AI video analysis module is deployed
on edge cloud and used in intelligent security, video monitoring, and face recognition
scenarios. Edge cloud can realize local analysis, rapid processing, and real-time
response due to its low latency, high bandwidth, and fast response, which effectively
make up for the disadvantages of AI-based video analysis, such as a high latency and
poor user experience.
Figure 5.4 Edge-Cloud–Based Intelligent Monitoring Solution
Figure 5.4 shows the AI video analysis system architecture based on the Edge-
Cloud service platform. The Edge-Cloud platform locally shunts the videos collected
by 4G cameras, which reduces the use of transmission bandwidth resources of the core
network and the backbone network. This shortens the end-to-end delay. The cloud
computing center performs AI training tasks, and the Edge-Cloud platform executes AI
inference. Local decisions and real-time responses make the system applicable to many
local AI applications, such as expression recognition, behavior detection, path tracking,
hot-spot management, and physical attribute recognition. In addition, cable cameras can
also access the intelligent Edge-Cloud analysis platform.
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
37 Copyright©China Unicom Network Technology Research Institute, 2018
Based on this solution, China Unicom has commercially deployed Edge-Cloud
Intelligent Security systems in Zhejiang and Hubei provinces of China in the Xueliang
(comprehensive security) projects. The system will be commercially used in more
provinces on a large scale.
5.4 Scenario 4: Augmented Reality (AR)
The biggest challenge of AR is the delay and bandwidth. To overcome the
challenge, AR information (such as user locations and camera angles) needs to be
locally processed to minimize the AR delay and improve the accuracy of data
processing. The existing AR solutions are generally based on the interaction between
mobile phones and the AR servers on the cloud. However, it always takes a long time
for image uploading, image recognition on the cloud, and information feedback due to
a large amount of image information. The long time will affect the user experience of
AR.
As shown in Figure 5.5, the AR server is deployed on the edge of network, the
Edge-Cloud platform determines the user's location through network data calculation.
The AR server provides real-time content aggregation and pushes the content,
shortening the delay of transmitting a large amount of image data, and effectively
reducing the image recognition and calculation load on the cloud. The content on the
edge server can be dynamically updated based on the characteristics of the region in
any scenario, so the user experience can be enhanced.
Figure 5.5 Edge-Cloud–Based AR Solution
Based on the above solution, China Unicom is carrying out pilot project of the
current network in scenic spots and museums. In a repair scenario, for example, the AR
glasses acquire and upload device images for the on-site engineer who wear the glasses.
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
38 Copyright©China Unicom Network Technology Research Institute, 2018
The repair app on the edge service platform identifies the images, and shows device
information and repair instructions on the AR glasses. Experts can remotely provide
written or audio guides on the back end. In a scenic spot scenario, the AR glasses take
pictures for a tourist and upload them. The value-added scenic app deployed on the
edge cloud platform identifies the images, displays enhanced scenic information on the
AR glasses, and navigates the tourist based on AR pictures. AR can be widely used in
numerous application scenarios, such as museums, art galleries, city memorial halls,
scenic spots, and factories, and greatly enhance user experience.
5.5 Scenario 5: Live VR Broadcast
In 2017, China further strengthened the strategic positioning of the VR industry,
and explicitly specifies VR as a development key in “The 13th Five-Year Plan for
National Science and Technology Innovation” and “The 13th Five-Year Plan for
National Strategic Emerging Industry Development”. Live VR broadcast differs from
traditional live broadcast in its high resolution ratio, high-code-rate 360-degree vision,
3D image and interaction. Considering the high-bandwidth and low-latency
requirements, live VR broadcast is greatly restricted in current networks.
The Edge-Cloud platform allows an ultra-low latency in live VR broadcast.
Traditional image collection cameras and VR cameras are wired. Using wireless cell
network backhaul instead of the wired solution is more helpful in mobile scenarios and
outdoor live broadcast. In addition, the distribution, assembling, and computation of
VR video and HD live-broadcast video can be unloaded to the network edge. Cameras
only collect image information, thus reducing costs. They can also be manufactured in
small size and easy to carry. The near-end distribution of live broadcast contents may
result in lower latency and better experience.
As shown in Figure 5.6, China Unicom, Intel, and Tecent Cloud built an Edge-
Cloud platform in the “Mercedes-Benz Cloud Center” situated in Shanghai, and
deployed the Edge-Cloud virtualized network function on the platform. The open API
and EVO is also provided. In this venue, the interaction between fans and web
celebrities can be interacted through the live VR broadcast or multi-angle HD video,
with an ultra-large bandwidth and a near-zero latency. Compared with the traditional
Internet (Center-Cloud) live broadcast with a latency of 30 to 40 seconds, the real-time
live broadcast greatly enhances user experience, gaining a good reputation among users.
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
39 Copyright©China Unicom Network Technology Research Institute, 2018
Figure 5.6 Live VR Broadcast Solution (Edge Cloud-Based)
Currently, as the official communication service partner of the 2022 Winter
Olympics, China Unicom has already begun the preparation and planning of over 20
Olympics venues located in Beijing, Yanqing, and Zhangjiakou. With the commercial
use of 5G networks, mobile live VR broadcast applications (Edge Cloud-based) will
surely reach a new vertex.
5.6 Scenario 6: IoV C-V2X
IoV refers to all-round network connections (inside vehicles, between vehicles and
men, between vehicles, between vehicles and roads, and between vehicles and service
platforms) by using a new generation of information and communication technologies
to enhance the intelligent level and automatic driving capabilities of vehicles. IoV will
construct new patterns of vehicles and transport services. This improves transportation
efficiency and enhances driving and riding experience, providing intelligent,
comfortable, safe, energy-saving, and efficient comprehensive services. As a new type
of application that develops very fast, IoV raises new requirements for service quality:
• Information and entertainment applications (environmental information
sharing and in-vehicle entertainment): a bandwidth of 10 to 100 Mbps and a latency of
500 ms.
• Traffic safety driving applications (platooning and anti-collision): a bandwidth
smaller than 1 Mbps, and a latency of 20 to 100 ms.
• Automatic driving: a bandwidth of 100 Mbps and a latency of 1 to 10 ms.
To meet the extreme requirements of IoV for latencies and reduce the load of the
bearer network and core network, the Edge Cloud-based IoV network architecture is
composed of an Edge-Cloud platform, RSU for drive testing, mobile terminals, vehicle-
mounted terminal (OBU), and traffic lights, see Figure 5.7. The RSU collects data
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
40 Copyright©China Unicom Network Technology Research Institute, 2018
uploaded by various terminals and distributes instructions to various terminals. The
Edge-Cloud platform can execute regional algorithms to implement safety avoidance,
speed guidance, and regional traffic analysis.The same time it can support handovers
and dynamic data synchronization for moving vehicles.
Figure 5.7 Edge Cloud-Based V2X Solution
Based on the above solution, China Unicom already starts the current-network
pilot site of V2X in Chongqing. The verified applications include the pushing of traffic
information based on relational models, APP development of driving vehicle queues
based on the Edge-Cloud Platform, intelligent traffic scheduling, and big data analysis.
• The real-time traffic information collected by the RSU for drive testing and
vehicle-mounted terminal (OBU) is sent back to the local Edge-Cloud server through
the LTE-V base station. The local Edge-Cloud server carries out the filtering,
comparison, association, and fitting operations based on relational models in
accordance with the sensing information in the specified region or area. And then
generates a real-time traffic information transfer chain. At last, it also pushes and
displays the traffic information about special scenarios (such as congestion, traffic
accident, and road information, as well as prealarm information) to in-vehicle terminals
or other terminals in accordance with the relational message model.
• By using location, route, planned arriving time, climate, and road data, the
Edge-Cloud platform carries out the filtering, computation, and combination operations
in accordance with the pre-configurations, dynamic requests, and route merging
information on the Edge-Cloud. It generates a proper vehicle list, route plan, driving
speed, and relevant platooning conditions, sends them to the corresponding vehicles,
and also provides decision information for vehicle-related servo organizations.
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
41 Copyright©China Unicom Network Technology Research Institute, 2018
• The intelligent traffic scheduling of the Edge-Cloud platform collects statistics
on vehicles and passengers in different directions in accordance with the V2X RSU-
collected traffic object information that is set at crossroads. The V2X-based traffic
object statistics and predication model is also designed, and analysis computation is
carried out on the Edge-Cloud platform. Traffic lights are dynamically adjusted in
accordance with the model to increase the traffic efficiency at crossroads.
• The big data and data analysis platform is used to construct a unified smart
core transportation system. The real-time status information, traffic and map data of the
IoV is analyzed, and data streams are sorted and classified. A model is then constructed,
so that a big data and data analysis platform can be constructed for IoV applications.
Finally, a unified smart core transportation system is built on the basis of this platform.
Chongqing is a city with complicated road condition in China. At the pilot site of
C-V2X, China Unicom selected the roads in Liangjiang New Area for experiment
purposes. 12 different types of typical roads (such as open and densely-populated roads
as well as roads with complicated traffic status, 9.6 km long) were covered, meeting the
complexity and typical requirements from the perspective of wireless coverage,
optimization, in-vehicle information collection, and Edge-Cloud data analysis. On the
experimental road, the prealarms about running red lights at signal crossroads, assistant
driving of signal-control crossroads, anti-collision reminding at crossroads, and vehicle
speed guidance at signal-control crossroads, emergency vehicle prealarms,
optimization of signal-light control parameters, road-crossing prealarms of passengers,
and simulation of platooning driving were tested. A great amount of rich and precious
experimental data was obtained, laying a foundation for the future IoV expansion
experiments and industrial applications.
5.7 Scenario 7: Indoor Positioning
Location information is the online identifies of real objects, and can truly reflect
the interaction between real worlds and virtual spaces in the three-dimensional
coordinate system. It is also the basis of IoE. However, the location problems of men
and things in IoT scenarios need to be solved first. However, isolated user location data
has very limited value. The business value of information can be fully expanded only
after the men, things, and data are connected and placed in scenarios through user
location data.
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
42 Copyright©China Unicom Network Technology Research Institute, 2018
The Edge-Cloud platform provides high-precision indoor positioning through
various access modes, and provides high-precision positioning services for smart
inventory and logistics, smart manufacturing, emergency rescue, personnel asset
management and service robots by using indoor high-precision positioning. Grid-level
low-precision positioning is provided for the retail industry. Location information
records the movement tracks and areas of each customer. If other big data (such as
consumption information, staying time, and behavior habits) is combined, disruptive
changes will be caused due to more precise marketing, more efficient delivery, more
precise pre-judgment analysis, and more efficient transaction.
Figure 5.8 Mixed Indoor Positioning Solution (Edge-Cloud Platform-Based)
As shown in Figure 5.8, China Unicom is committed to launching a smart
shopping mall solution based on the Edge-Cloud platform. This solution involves
multiple positioning technologies such as indoor base stations, WiFi, and NB-IoT,
allowing five-meters high-precision positioning. At the same time, the Edge-Cloud-
based IoT management platform can manage wireless sensors jointly such as
geomagnetic field sensors, fire sprinklers, and fire alarms. In addition, the Edge-Cloud
platform is used to provide the capabilities of indoor location services for third-party
applications and the big data platform of shopping malls. Indoor navigation and smart
parking are provided for users, and property management is also provided for owners.
Indoor locations and big data are combined to excavate valid data for further analysis
and calculation, aiming to get the highest valuable information.
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
43 Copyright©China Unicom Network Technology Research Institute, 2018
5.8 Scenario 8: Industrial IoT
Currently, smart manufacturing, electricity, and public utilities have been digitized,
intelligized, and network-oriented, and face a series of key challenges.The challenges
includes: OT and ICT cross-border collaboration, interconnection and integration of
physical worlds and digital worlds, valid information flowing and integration caused
by the co-existence of multiple heterogeneous technologies, integration of industrial
Know-How and data driving in the process of knowledge modelization, and end-to-end
collaboration and integration resulting from the extension of the industrial chain.
Based on the edge computing model and intelligent distributed architecture, an
intelligent Edge-Cloud Platform can be constructed on the industrial site to achieve
unified network connections, unified intelligent distributed architecture, unified
information models, unified data services, unified control models, and unified service
orchestration. The industrial Edge-Cloud integrates local network, computation,
storage, and application core capabilities inside factories, and provides a nearby local
cloud platform for production management services. It complies with the mode of
“Moving Factory Management to the Edge End and Moving Perceiving End Upwards”,
and conditionally integrates OT, IT, and CT. In this way, the massive, heterogeneous,
and real-time connections between terminals and devices are achieved, and networks
are automatically deployed and maintained to ensure the safety of connections. The
hardware infrastructure of industrial Edge-Cloud uses a universal X86 architecture
server, and the industrial characteristics are stronger than those of the commercial server,
providing more stability and security. Based on the Openstack or Kubernets technology,
a VM-based or Container-based virtualized platform is built on the basis of hardware
infrastructure, aiming to placing computation, storage, and network resources in a pool.
This allows application developers in vertical industries to design various applications.
Figure 5.9 shows the function architecture of the industrial Edge-Cloud Platform.
This architecture provides a development and deployment operation service framework,
and can achieve intelligent collaboration between development and deployment. In this
way, the unification of software development interfaces and the automation of
deployment operation are implemented. The southbound interfaces can interconnect
industrial sensors and devices. The northbound interfaces can interconnect industrial
APPs and implement various typical applications through OT&ICT integration,
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
44 Copyright©China Unicom Network Technology Research Institute, 2018
including the predictive maintenance of industrial devices, power consumption analysis
of industrial sites, service orchestration of the industrial production process, and
network connection of devices on industrial sites. Additionally, the industrial Edge-
Cloud Platform allows multiple types of vertical applications customized for different
production enterprises, including the online real-time status detection of production
devices, online maintenance of production devices, network connection of the edge
cloud platform, and remote cloud-end management.
Figure 5.9 Function Architecture of the Industrial Edge-Cloud Platform
Currently, China Unicom positively responds to the national “Internet +
Intelligent Manufacturing” strategy, and works along with the Guangdong Intelligent
Manufacturing Institute and South China Intelligent Robot Innovative Institute for
intelligent manufacturing Edge-Cloud pilot engineering held in Foshan, Guangdong
province. Besides, China Unicom has launched the R&D and service demonstration
project towards the 5G network-oriented industrial manufacturing Edge-Cloud
Platform in accordance with the key national technological project “Industrial
Manufacturing-Oriented 5G Service R&D and Experiments”.
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
45 Copyright©China Unicom Network Technology Research Institute, 2018
6 Summary and Forecasting
When facing the coming 5G era and successive tides of ICT integration, we are
ready to brace the future through a series of reforms. Different from the role of pipeline
provider in the 4G era, telecom operators will gain more opportunies to expand their
value-added services in the coming 5G era, and become comprehensive end-to-end
service providers. As a new technology of ICT integration, edge computing deploys
high-bandwidth, low-latency, and localized services onto the network edge, and
provides unified telecom infrastructure support for FMC, which is of vital importance
for telecom operator’s digital transformation and industrial structure upgrade. China
Unicom's Edge-Cloud platform will provide differentiated network capabilities and
services for application parties in vertical industries, and provide unified operation
management support, thus significantly enhancing industrial competitiveness.
Although both standards organizations and industry alliances are actively engaged
in the research of the technological development of edge computing, constructing the
ecology of the edge industry needs more open collaborative research and bold industrial
practice. By working with Baidu, Alibaba, Tecent, ZTE, and Intel, China Unicom has
begun large-scale Edge-Cloud pilot projects and pre-commerical network construction
in 15 provinces and cities of China. It is committed to building an open and open-source
Edge-Cloud Service PaaS Platform to flexible allocate computation, storage, network,
and accelerator resources, carry out orchestration amangement for varied edge services,
and provide rich platform service capabilities and unified APIs for application
developers, aiming to accelerate the incubation and promotion of edge services. This
successful industrial practice will be of great guidance for the construction of the
ecology of edge computing. Next, China Unicom will pursue the exploration of
nationwide smart campuses, smart venues, IoV V2X, and industrial IoT.
All industries have unlimited anticipation and expectation for various application
scenarios of edge service platforms in the future. However, the fulfillment of this dream
needs the joint effort of the whole industrial chain. China Unicom is expecting to work
with more industrial partners during the next large-scale Edge-Cloud pilot projects and
the advancement of commercial use, and jointly discuss the collaboration modes of
edge service platforms, aiming to co-build a 5G-oriented edge ecosystem and fully
promote the vigorous development of edge services.
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
46 Copyright©China Unicom Network Technology Research Institute, 2018
This White Paper is Co-Written by
• ZTE INTEL
China Unicom's Edge-Cloud Ecological Partners
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
47 Copyright©China Unicom Network Technology Research Institute, 2018
Abbreviations
Abbreviation Full Name
OT Operation Technology
IT Information Technology
CT Communication Techonology
ICT Information and Communication Technology
MEC Multi-access Edge Computing
OTT Over The Top
CDN Content Delivery Network
DC Data Center
eMBB enhanced Mobile Broadband
mMTC massive Machine Type of Communication
uRLLC ultra Reliable & Low Latency Communication
CAPEX Capital Expenditure
OPEX Operating Expense
CUPS Control and User Plane Separation of EPC nodes
HGU Home Gateway Unit
OLT Optical Line Terminal
BRAS Broadband Remote Access Server
SDN Software Defined Network
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
48 Copyright©China Unicom Network Technology Research Institute, 2018
CORD Central Office Re-architected as a Data Center
NFV Network Function Virtualization
NFVI Network Functions Virtualisation Infrastructure
NFVO Network Functions Virtualisation Orchestrator
VNF Virtual network function
VNFM VNF Manager
MANO MANagement and Orchestration
DevOps Development and Operations
UPF User Plane Function
RNIS Radio Network Information Service
CPE Customer Premise Equipment
OSS Operation support system
AMF Access and Mobility Management Function
SMF Session Management Function
MME Mobility Management Entity
IoT Internet of Things
NB-IoT Narrow Band Internet of Things
CU Central Unit
DU Distributed Unit
SBC Session Border Controller
BGN Broadband Network Gateway
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
49 Copyright©China Unicom Network Technology Research Institute, 2018
APP Application
API Application Programming Interface
IDC Internet Data Center
COTS Commercial Off The Shelf
BIOS Basic Input Output System
FW FirmWare
OS Operation System
GPU Graphics Processing Unit
FPGA Field Programmable Gate Array
VIM Virtualised Infrastructure Manager
PIM Physical Infrastructure Manager
CMP Cloud management platforms
MEPO Multi-access Edge Platform Operation
MEPM Multi-access Edge Platform Manager
RO Resource Orchestrator
NS Network Service
NSO Network Service Orchestrator
LCM Life Cycle Management
SRS Sounding Reference Signal
RRU Radio Remote Unit
GTP GPRS Tunnelling Protocol
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
50 Copyright©China Unicom Network Technology Research Institute, 2018
AR Augmented Reality
VR Virtual Reality
CQI Channel Quality Indicator
SINR Signal to Interference plus Noise Ratio
BLER Block Error Ratio
QoS Quality of Service
MQTT Message Queuing Telemetry Transport
COAP Constrained Application Protocol
FTTP Fiber To The Premise
FTTC Fibre To The Cabinet
V2X Vehicle to Everything
NAPT Network Address Port Translation
AI Artificial Intelligence
EVO EDGE Video Orchestration
RSU Road Side Unit
OBU On Board Unit
White Paper for China Unicom's Edge-Cloud Platform Architecture and Industrial Eco-System
51 Copyright©China Unicom Network Technology Research Institute, 2018